Rutgers University

ECE 332:418

Retired Lithium Ion Electric

Vehicle Batteries Making the Most Out of their Second Use

Jarek Roszko, Mohammad Khan, Ammar Sal, and Nabil Ali

Table of Contents

Page #

1. Table of Contents 2

2. Introduction (Background and Contributions) 3

3. Technical Background (Hardware) 3

4. Schematic 7

5. Technical Background (Software) 8

6. Experimental Results 16

7. Project Enhancement and Future Work 18

8. Economic analysis and Impact 18

9. Conclusion 18

10. Acknowledgements 19

11. References 20

Introduction

Lithium ion batteries are some of the most widely used rechargeable batteries in the world. They are

present in portable devices such as cell phones, handheld video game consoles, mp3 players, laptops, cameras,

etc. They are even used for some power tools. Virtually any device that relies on a rechargeable power source

uses lithium ion batteries. It is daunting to imagine a society without lithium ion batteries. Purchasing traditional

one time use batteries repeatedly in a vicious cycle both generates more waste and forces consumers to spend

more on energy.

Because of lithium ion, the practicality of electric cars such as Tesla and Chevy Volt was attractive.

Electric vehicles (EV’s) and plug-in hybrid electric vehicles (PHEV’s) are gaining popularity in the US and

around the world because they are promoted as environmentally friendly cars. Air pollution, record high gas

prices, and dependence on foreign oil are pushing sales growth of the EV’s and PHEV’s. Advertisements assure

us of “zero emission” and the question asked is no longer “why electric?”, but “why gasoline?” While most

electric car owners consider themselves “green”, the process of mining for lithium and the production of these

batteries in reality are actually not as green as they think. In addition, the recycling process is neither simple nor

cheap. Our objective for this capstone design is to utilize retired lithium ion batteries from EV’s and PHEV’s as

a reusable power source for residential applications.

We have divided the project into 2 phases: hardware and software. In the first phase, Ammar Saleem

and Jarek Roszko worked on the hardware layout of the project while in the second phase, Mohammad Khan

and Nabil Ali measured and monitored the core component of the project: the battery bank. Dividing the

workload and focus in this way was efficient. It will be efficient to work in group of two people and

successfully complete the project on time. Also, it will be very helpful to troubleshoot if any error occurs while

testing.

Hardware

The hardware phase consists of planning the layout of the physical electronic materials, testing and

creating the lithium ion battery banks and the load circuit. We created our battery bank with a 12 V maximum

output consisting of three lithium ion cells. Each cell current specification is at 2 Amps. Being wary of the

dangerous volatility of lithium ion batteries, we are installing several safety parameters to our project.

The most important safety device that we acquired was a battery charger. We had ordered the Intelligent

Digital Balance Charger/Discharger for (NiMH/NiCD 1.2V-18V, Li-Ion 7.4V-22.2V, LFE 6.4V-19.2V, SLA

2V-20V) battery packs and found it to be the best way to safely charge the battery cells at the appropriate rates.

Balanced charging was a crucial component in our project, and if we had been unable to obtain the battery

charger, our project would have been much more difficult. We would have needed to find a way to keep the rate

of charge appropriate for each rate simultaneously, a task that would have been beyond the scope of this 3

month semester project. Though the National Instruments myDAQ helped us take measurements from 4 ports,

we would have needed double the amount to measure the voltages of each cell in unison. Regardless, the

myDAQ did not have balancing capabilities and the need for safety parameters would have been significantly

larger. This need for balancing was biggest obstacle we had throughout our project. Because of the battery

charger, we were able to overcome our difficulties and land one step closer to our goal.

As a failsafe, we installed a circuit breaker and temperature sensors with automatic cutoff options onto

the battery cells. The temperature sensors were a remarkable addition in that they ceased the charger from

supplying power to the cells if temperatures exceeded 60 degrees Celsius, a threshold which conveniently

reflected the lithium ion battery operating parameters. This was a key piece of hardware that assured us that in

the case of a power surge, where the battery charger would have overloaded and supercharged the batteries, the

temperature sensors would have broken the wire connections from the charger and the bank. This is what would

have prevented the lithium ion batteries from being overcharged and thus, self-incinerating.

To emulate a residential home, we assembled a small scale load circuit to test the battery bank’s

functionality. In the load circuit, we included three light bulbs and a small fan. We also included a fan to

provide an additional source of cooling for the battery bank.

Figure (1) The Battery Banks emulating a laptop charger

Figure (2) The Load Circuit with lightbulbs and fan

Though we did not use myDAQ to balance our battery bank, it was an essential tool for measuring the

voltage and current of our battery bank. In addition, we constructed a small scale simple RC circuit to learn the

proper experimental measurement procedure before testing the lithium ion battery cells. By testing for voltage

and current on the RC circuit, we were able to better learn about the myDAQ and its LabView partner software.

It was crucial for us to have a standardized testing procedure on the RC circuit that we could then carry over to

the actual battery bank and load circuit. If we had not refined our testing procedure and jumped straight into

measuring and experimenting on the actual project circuit, we would have been in trouble if we ran into

unforseen complications. Because we are dealing with lithium ion batteries, it is much better to run into those

“unforseen complications” on a simple RC circuit than on the real deal load circuit where in the latter, bulbs can

burst and fires can occur.

Figure (3a) Above, the RC Circuit with Resistor and Capacitor

Figure (3b) Below, the myDAQ supplying power and taking measurements

Figure (4) Schematic

Software

Throughout the Information age, the rise of software became more and more prominent. Today, many

hardware systems have complementing software systems to guide them. In a way, a hardware and software

system can be seen as a metaphor for the human body and brain. The brain is what guides the body, tells it what

to do, etc. The body is heavily reliant on the brain’s health. Without the brain, the body is just a shell.

Analogous to many systems, the hardware is heavily reliant on the software. Without a prominent software

component, the hardware component cannot fulfill the same level of performance. Let’s take a vehicle

manufacturing facility. The hardware is the mechanical arms that piece together a car from different parts. With

a human operator, comes the concern over stamina, precision, and efficiency. But with a software component

that takes the human operator’s role, you have unlimited stamina, maximum precision, low cost, and an overall

efficient system.

Though the hardware component of our project was very important, the complementing software side

was equally as important. Due to the dangerous volatility of lithium ion batteries, it is important to be wary of

the batteries’ voltages so that they are not overcharged or too discharged.

To observe and maintain the batteries’ voltages we are using the National Instrument’s myDAQ.

Connected with NI’s LabView software, we are able to observe and record data measurements ranging from

current to voltage.

Without LabView and the myDAQ, we would have been unable to take measurements and plot real time

graphs to show the charge and discharge of the lithium ion battery bank.

Figure (5) The myDAQ Hardware and LabView Software

http://images.studica.com/images/product/

National-Instruments-Mini-Systems-

Accessories/94myVTOL%20miniSystem%20

07161207.jpg

We had several goals to accomplish from a software perspective. We first used myDAQ and LabView to

measure and test the RC circuit board. It is important to emulate our procedure on a low risk circuit board so

that we may perfect our testing procedure on the actual lithium ion board design. The RC circuit board provides

room for errors that we can avoid for the dangerous lithium ion batteries.

Figure (6) Battery voltage measurement example

We have successfully taken voltage and current measurements from the RC circuit using the myDAQ

and plotted our data with LabView. We created a LabView program to automate the measurement process. To

record documented data and efficient graphs, we learned how to export data to Microsoft excel.

Figure (7) Our LabView program, with Excel charts and graphs

http://www.ni.com/cms/images/devzone/t

ut/measure_battery.PNG

Fortunately, LabView is a very in depth software that provides users with many options and tools for

common experimental needs. However, one challenge that we were trying to overcome is being able to stop

power from supplying into the RC circuit board. This is a very crucial ability that we as the software designers

need to be able to take advantage of. The usefulness of this feature will be most noticeable when experimenting

with the lithium ion battery circuit. If we notice that batteries approach the minimum discharge voltage, we

want to cease the process. Alternatively, if we notice that the batteries approach the maximum charge level, we

also want to maybe use the myDAQ to cease charging the batteries. This is important because lithium ion

batteries can’t be too overcharged or too discharged. If we transgress beyond a certain threshold, cataclysmic

damage can occur and the batteries will self-ignite.

We found that the best way to tackle this dilemma was through the battery charger. Initially, we were

very concerned with balancing the lithium ion cells. Each of the 6 battery cells had different voltages. If we

were to charge them all at once at the same rate, some cells would have been overcharged, and this would have

caused an ignition that would have caused a fire to spread to the other batteries and the whole design board. As

our project progressed, we were able to get our hands on the battery charger that relieved us of having to worry

about the balancing dilemma from a software and control point of view.

Another challenge we faced was how to be able to use multiple myDAQs to take multiple voltage

measurements. Fortunately, LabView is powerful software that can support multiple myDAQ utilizations from

the same computer. Since we did not have to worry about the balancing of our battery bank, our main focus

became measuring and logging in the voltage and current at the terminals of the bank(s) just after the fuse to the

ground. Therefore, to measure voltage of the bank, we used the DMM ports of the myDAQ to measure the

analog voltage with respect to time and simultaneously to log it into a technical data management system

(TDMS) file or excel document. All the data acquisition was done in 500 milliseconds sampling rate through all

the ports or channels. MyDAQ was perfect for data acquisition since our measurement needed a minute time

scale instead of smaller time divisions such as nanoseconds or milliseconds. The MyDAQ can acquire data as

fast as 200K samples per second as per the manufacturer’s rating. Since our main focus was to acquire current

and voltage readings, we had two myDAQ assistants in our software. We needed to acquire this data for our

State of Charge (SOC) or Depth of Discharge (DOD) and State of Health (SOH) calculations. In our case,

voltage translation was used to estimate the SOC usually focusing on the initial and final parts of the discharge

curve where we had more change, since the central portion of the curve is almost constant. Knowing the

relationship of open-circuit battery voltage and SOC (which is the idea behind voltage translation) allows the

voltmeter to be calibrated to report an approximate SOC. The software measured Voltage with time to acquire

estimated DOD when the cell voltage crosses a threshold which in our case was 9 V for the module voltage.

The threshold value was acquired by manufacturer’s datasheet. The software handled this case by using simple

comparator that directly correlates with the readings from the measured voltage matrix within our state machine

and constantly beeps which tells the user (in manual mode) to turn on the charging balancer for the battery

packs. In the real world, this will be automated in conjunction to the charging platform, where the control

system such a simple PLC will manage the charging platform, load, and as well monitor the life indication of

the pack to the end user.

One of the greatest challenges was to acquire the current through the load center without shorting the

load or destroying the actual measuring device. There are two ways to measure the current in such cases; such

methods are half effect transducer or current shunt method. We chose half effect method since the current shunt

provides some energy / power losses, hard to implement in the hardware, and the resistance of current shunt

changes with temperature. A Hall Effect sensor is placed inside the magnetic field produced by a cable that

carries the pack current. It produces a voltage that is proportional to that current; that voltage can be measured

directly (Figure 8a). High-current Hall effect sensors are modules shaped like a ring, through whose opening a

cable carrying the pack current is routed. Low-current Hall effect sensors are ICs with two power terminals,

through which the current is routed. Hall Effect sensors are characterized by the following:

“The current reported by a Hall Effect sensor remains accurate over time and temperature.”

Hall Effect sensors are isolated from the pack current and therefore no isolation is needed. Hall Effect

sensors suffer from offset at 0 current, which changes with temperature. So, even if they are zeroed at room

temperature, they will report a small current when there isn’t one as they get hot or cold. Frequent calibration is

possible in applications that have periods of 0 current, such as HEVs. Hall Effect current sensors are modules

that include their own amplifier, so, unlike the signal of current shuts, their output is at a high level. They can be

powered by one supply (5V) or two supplies (+/— 12V or +/— 15V), and they can be unidirectional (can only

see current in one direction) or bidirectional (can see both charging and discharging current). Based on that,

their output can be referenced to ground (0V at 0A), or have an offset (typically 2.5V at 0A). In particular, the

output of a two-supply bidirectional sensor is bipolar—it will swing above and below ground. The analog input

on a BMS needs to be compatible with the output voltage of the current sensor: 0- to 5-V output, or −12V to

+12V. To make a bipolar current sensor work with a 0- to 5-V input, a 2:1 voltage divider is required, with one

resistor in series with the signal and the other resistor between the BMS input and a 5-V supply (Andrea, 52-

53).

Figure 8a: Hall Effect current sensor circuit Figure 8b: Current transducer (Model: HX-10P)

In our case the current transducer detects electrical current at the load in a wire, and generates an analog

signal proportional to it, which is read by myDAQ analog input in terms of voltage. The output voltage reading

from the Hall Effect had to be calibrated in myDAQ to give us the actual current flowing through the load

center. Calibration of the Hall Effect transducer was done by measuring the voltage drop at load through DMM

and the voltage output from sensor in myDAQ each time by varying the load at the load center. Since P=IV, we

calculated Ii=Paladin/Vload_i. This data of Ii@load_i versus Vcurrent sensor output was then plotted on excel. After we got

the v(i) function from the data, it was a linear curve with the equation v(i) = 0.4027i-0.0046. To implement this

into our myDAQ we have to find inverse function i(vsensor_Output) to get the actual current measured by the

current sensor which gave us isensor(v) = (vsensor_Output+0.0046)/(0.4027). This was then implemented in myDAQ

by taking the sensor analog voltage output and scaling it to current reading which was graphed and logged

simultaneously. Knowledge of the battery current allows a BMS to perform additional functions, which, while

not essential, are expected to be offered by a professional product. These are, in order of likelihood that a

particular BMS will implement that function which prevents the cells in the battery from being operated outside

their safe operating area (SOA) in terms of continuous current (analog BMSs that measure battery current

usually implements just this one point).

Now that we successfully know how to acquire our load terminal voltage and current we need to

implement this with myDAQ using the LabView software. Our data acquisition was implemented in LabView

using a state machine. The state machine is one of the engineering fundamentals frequently used to build

applications quickly. Developers use state machines in applications where distinguishable states exist. Each

state can lead to one or multiple states and can end the process flow. A state machine relies on user input or in-

state calculation to determine which state to go to next. Many applications require an “initialize” state followed

by a default state, where you can perform many different actions. These actions depend on previous and current

inputs as well as states. You can use a “shutdown” state to perform cleanup actions. When designing state

machines, you can create a state diagram to graphically represent the different states and how they interact. Use

state diagrams, the design frameworks for state machines, to model the control algorithms you need with

discrete logical states. State Diagrams make it easy to develop and understand the functionality of an

application that uses a state machine. In our application we used flow chart to represent the functionality of our

state machine. Figure 8a illustrates the flow chart and the state machine is illustrated in Figure 8b. We placed a

while loop on the block diagram. The state machine was built on the basis of placing a case structure in the

while loop then creating a shift register on the while loop where we then created an Enum and wired it to the

shift register to initialize it. We stop and log the data matrix when we go outside the while loop once the display

history is blocked or the stop button has been activated (clicked by the user) in the GUI interface as seen in

figure 10.

Figure 9a: LabView BMS software basic flow chart

Figure 9b: Data Acquisition State Machine

Figure 10: GUI for the BMS SOH estimation showing discharging voltage and current discharge curves

Now that we got the data from myDAQ in a TDMS file/ excel file, we can graph other components of

discharge curve to give us more insight of the battery bank. Such curves such as state of charge (SOC), depth of

discharge (DOD), and Voltage vs. DOD were necessary to convey the health of the battery banks. The state of

charge (SOC) of a cell or a battery at a given time is the proportion of the charge available at that point,

compared to the total charge available when it is fully charged. It is expressed in percent, from 100% when full,

to 0% when empty. The SOC evaluation function is also known as the fuel gauge, especially in EVs, because of

its analogy to a gas car’s fuel gauge. It is essential to understand that each cell in a battery has its own SOC, and

that the battery itself has its own, separate SOC. The depth of discharge (DOD) of a cell or battery is a measure

of the charge removed from it. It is expressed in amp-hour (Ah). DOD can be expressed in a percentage as well,

and is commonly done so in Lead Acid batteries. It is really more useful to express DOD in Ah, so that the

combination of SOC (in percent) and the DOD (in Ah) conveys more information than would be the case if both

were expressed in percentages (Andrea, 25). The graphs pertaining to discharging of the banks are listed on the

next page.

-0.2

0

0.2

0.4

0.6

0.8

0 0.005 0.01 0.015 0.02 0.025

Cu

rren

t (A

)

Time (hours)Fig. 11.1

Current vs. Time while discharge

0

2

4

6

8

10

12

14

0 0.005 0.01 0.015 0.02 0.025

Vo

ltag

e (V

)

Time (Hours)Fig. 11.2

Voltage vs time during discharge

0

2

4

6

8

10

12

14

-0.005 0 0.005 0.01 0.015

Vo

ltag

e (V

)

Discharge Capacity (Ah)Fig. 11.3

Voltage vs Discharge capacity during discharge

15.5

16

16.5

17

17.5

18

18.5

0 20 40 60 80

Res

ista

nce

(O

hm

s)

Time (Seconds)Fig. 11.4

Estimated Resistance vs Time

y = -1.2492x + 106.03

-20

0

20

40

60

80

100

120

0 20 40 60 80 100

SOC

(%

)

Time (Seconds)Fig. 11.5

estimated SOC vs time

y = 0.5833x + 2E-05

-0.002

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0 0.01 0.02 0.03

Do

D (

Ah

)

Time (Hours)Fig. 11.6

Depth of Discharge vs Time during discharge

The current and voltage versus time, figure 11.1 and 11.2 respectively was acquired by measuring them

from load using myDAQ from a log file in excel. Figure 11.3 shows us the rate of change of bank resistance

over time, where we have decrease in current and voltage. The nominal voltage of a galvanic cell is fixed by the

electrochemical characteristics of the active chemicals used in the cell, the so called cell chemistry. The actual

voltage appearing at the terminals at any particular time, as with any cell, depends on the load current and the

internal impedance of the cell and this varies with, temperature, the state of charge and with the age of the cell.

The resistance of a battery provides useful information about its performance and detects hidden trouble spots.

High resistance values are often the triggering point to replace an aging battery, and determining resistance is

especially useful in checking stationary batteries. Measuring the internal resistance is done by reading the

voltage drop on a load current or by AC impedance. The results are in ohmic values. There is a notion that

internal resistance is related to capacity, and this is false. The resistance of many batteries stays flat through

most of the service life. In our case the estimated resistance is non-linear which is out of ordinary. This tells us

that our pack has a big drop in performance starting 40 seconds to later time. In our case the discharge curve fell

fast after 40 seconds, which tells us that internal resistance of the cells are large, and are close to being “dead or

empty” since it hit the cut in voltage of 9 V within 40 seconds. The resistance curve is shown in figure 11.4

where you can clearly see the sudden change in internal resistance while discharging. Figure 11.6 shows us

depth of discharge curve with respect to time, which tells us the fact how much charge is being removed over

time from the battery bank. In our case, the DOD curve slope was very steep with a slope of 0.583 A of charge

rate every hour, which means in terms of 2 A delivery, our battery will dry out or over discharge in several

minutes. Figure 11.5 displays SOC in percentage versus time that clearly tells us the rate of percentage of

charge from total capacity has been removed from the battery bank. In our testing procedure, the slope of SOC

versus time was about a 1.25 decrease in the state of charge while discharging. This is a very steep slope, just

over a matter of 84 seconds, therefore our battery banks are very old and used. This rough indication of battery

capacity over time tells us that these batteries are old and have built high internal impedance over time,

therefore they need to be replaced.

Project Enhancement and Future Work

Due to a low budget and time constraints of the project, we were not able to implement temperature

effects on the pack. This plays an important role in the characteristics of these SOH curves and the lithium cell

chemistry overall. In terms of software, we also wished to include a virtual button automation of the system

with LabView (using myDAQ’s DI/DO channels with a transistor logic system). With the buttons, we would be

able to choose the banks we wanted to use instead of flipping the switches. In addition, we could have also

implemented a better graphical interface that stated the state of health in terms of percentage while myDAQ

constantly monitored the packs. This project can be made very vast in terms of user end interface, automation,

and smart control system for balancing purposes if needed. On the other hand, our BMS was successfully

designed in terms of hardware and software. The goal of this project that was to measure the state of health of

our battery packs, provide a balancing mechanism, protection, data acquisition and telemetry, and most

importantly, convince the user and society that this model can be used in a commercial and household energy

system when needed.

Economic Analysis and Impact

There are lots of economical advantages of using retired lithium batteries from electric cars for

residential purpose. First, it will help promote green energy and make good usage of retired lithium batteries.

Second, it can help countries such as India, Pakistan, Sri Lanka and Bangladesh. Third, it will give

Manufacturers of Electric Car to earn profit from their retired Lithium batteries that cost up to $30,000 to

manufacture. Forth, it will reduce the dependability on electricity from Nuclear power plant.

Conclusion

In this project we have successfully proposed a practical circuit-based model for the state of health

(SOF) estimation of lithium ion batteries from electric cars (EV’s) for residential use. It was said above that

SOH is a measure of the batteries ability to store and deliver electrical charge. There are few methods out there

that can be applied to measure the SOH but we have used the most common one based on the battery capacity

C. A simple formula for SOH = (C_old/C_original) *100% was implied; which simply divides the old

capacitance of a battery measurement in time over the original battery capacity provided by the manufacturer as

a nominal rating. In other words the estimations are based on the Ampere-hour throughput (Ah). The Li-ion

cells used in our experiment are rated by the manufacturer at 2 Ah which is our C_original in a formula. Few

discharge experiment were performed using constant currant load of 2 Amps as seen in Figure 8. Our battery

banks maintained almost a linear behavior for only about 5 minutes and the sharp voltage drop is observed

yielding [(2 A * 5min)/(2A *60min)] * 100% gives us only about 10% of life left in the bank. The Li-ion cells

that make up battery bank are at the end of their useful life and can be safely recycled. Even though our

experiments were successful the circuit model used was with a small battery bank and the validity of the real-

life banks ranging from 350~400 Volts DC would have to be still confirmed. Testing such large capacity packs

would require a sophisticated Battery Management System for balancing due to series connected Li-ion cells

almost all the time.

Special Thanks

We would like to thank Professors Hana Godrich

and Yicheng Lu for their steadfast counseling and support

References:

1."National Instruments: Test, Measurement, and Embedded Systems." National Instruments: Test,

Measurement, and Embedded Systems. N.p., n.d. Web. 28 Mar. 2014

2. Andrea, Davide. Battery Management Systems for Large Lithium-ion Battery Packs. Boston: Artech House,

2010. Print.

3. http://images.studica.com/images/product/National-Instruments-Mini-Systems-

Accessories/94myVTOL%20miniSystem%2007161207.jpg

4. http://www.ni.com/cms/images/devzone/tut/measure_battery.PNG